energies

Article

Experiment on Bidirectional Single Phase ConverterApplying Model Predictive Current Controller

Gabriele D’Antona 1, Roberto Faranda 1, Hossein Hafezi 1,* and Marco Bugliesi 2

1 Politecnico di Milano, 20156 Milano, Italy; gabriele.dantona@polimi.it (G.D.);roberto.faranda@polimi.it (R.F.)

2 Studio Progettazione Informatica Bugliesi (SpiB.it), 20877 Roncello, Italy; bugliesi@spib.it* Correspondence: hossein.hafezi@polimi.it; Tel.: +39-02-2399-8523

Academic Editor: Rodolfo AraneoReceived: 14 January 2016; Accepted: 14 March 2016; Published: 24 March 2016

Abstract: A bidirectional converter able to manage storage is a basic power electronics device, andit is a major component of renewable energy sources, micro grid and also the smart grid concept.In this paper, single-phase bidirectional converter topology is discussed. The state space model hasbeen derived, and a simple model based predictive current controller has been utilized to controlthe inverter. Control block diagrams have been designed with MATLAB and simulation results arepresented and compared with experimental ones, giving credibility to the derived model and thedesigned control method.

Keywords: single-phase converter; model predictive control; Pulse Width Modulation (PWM);micro grid; smart grid

1. Introduction

Increasing global warming and energy consumption demand made it unavoidable to look for newenergy sources as alternative to conventional fossil energies [1]. Green and Renewable Energy Sources(RES) have attracted considerable attention in recent decades due to their clean and sustainable nature,in which power electronics play an important role in this road [1]. However, RESs spread aroundand are available locally. Thus, these Distributed energy Resources (DRs) should be tackled locallywherever those are available by utilizing the proper device. Thus far, DRs and Micro Grids (MG) havecome into the picture and are desirable for smart grid functions [2,3]. These applications mostly needproper integration of RES into the grid.

The main concept of integrating RES into the grid and implementing within smart grid systemsis to use a bidirectional converter that is able to manage a storage system [3,4]. However, becauseof intrinsic characteristics of inverters, like low-impedance and quick response [5], their current andvoltage profile changes, rapidly, therefore, within a fraction of cycle, those may overload, exceed thecurrent limit and cause failure of the system [5,6]. Thus, capability of the systems is affected by thecontrol and performance of the inverter.

The control of the inverter is an important issue in these kinds of applications because highquality power is desired [7]. Bidirectional inverters are the main parts of applications like DistributedGeneration (DG), energy-storage systems, standalone application based on RES and UninterruptiblePower Supplies (UPSs) [8,9].

Several reliable and effective control methods based on classical control algorithms with PulseWidth Modulation (PWM), like conventional Voltage Oriented Control (VOC), Direct Power Control(DPC) [10,11], deadbeat control [12,13], multiloop feedback control [14,15], adaptive control based onbank resonant filters [16], and repetitive-based controllers [17] have been introduced and implementedto industrial applications.

Energies 2016, 9, 233; doi:10.3390/en9040233 www.mdpi.com/journal/energies

Energies 2016, 9, 233 2 of 14

Recent fast development in Microprocessors, Digital Signal Processors (DSP) andField-Programmable Gate Array (FPGA) has affected power electronics devices deeply. Researchersare continuously looking for fast and more reliable control schemes for complex power electronicdevices. Model Predictive Control (MPC) has now roughly three decades of development [10,18], andit is considered one of the biggest advances in converter controls recently [10,19].

Conceptually, MPC can be any control method that predicts the future state of the system by usinga model of it, and selection criteria can be involved in optimization functions or not [19]. It appears asan attractive alternative for power electronics applications due to its fast dynamic response.

MPC offers many superior features and can be used in a wide range of applications and processcontrol. It is highly compatible to be implemented in multivariable systems and has been consideredone of the best solutions for the nonlinear processes [11,20]. MPC is a very flexible control method fornonlinear systems, and it is highly compatible to include in system constraints [7].

It has been implemented in multilevel cascade H-Bridge Inverter [21,22], Matrix Converter [23,24],UPS [7], and many other wide range power electronic applications [20]. However, it has somedrawbacks that should be addressed and compensated [20,25,26].

As reported in several recent works, although MPC has superior performance in transientresponse, decoupled control and flexibility, it has some drawbacks such as variable switching frequencyand wide spectrum. Control methods based on classic Proportional Integral (PI) controllers with PWMhave mostly presented a better steady-state performance. Thus, a number of recent works have aimedto deal with these drawbacks and improve the control strategy [20]. However, most works have useddiscrete model (often for three phase system) with optimization cost function [7,11,15,20–23]. Dueto the nature of MPC, its integration into other types of controllers like PI to improve the inverterperformance is difficult.

This paper presents a continuous model and a novel Model based Predictive Current Controller(MPCC) as a current controller a for bidirectional single-phase converter, which is able to manage astorage system. In the voltage source mode, a combination of PI and MPCC (PI-MPCC) is proposed,and advantages of both conventional PI controller and MPCC are utilized to enhance performance andreliability of the system. Simulation and experimental results are presented and compared.

The rest of the paper is organized as follows: Section 2 shows the state space model, Section 3presents control block diagrams, in Section 4, simulation and experimental results are shown, and,finally, conclusions have been made in Section 5.

2. System Model

Schematic of the implemented bidirectional single phase converter is shown in Figure 1a, and theexperimental prototype is shown in Figure 1b. The schematic from top to down, includes the AC sideinductance 1©; the full bridge Voltage Source Inverter (VSI) with four controlled switches as DC toac interface 2©; the DC bus capacitor 3©; the DC chopper leg with two controlled switches in order tomanage storage 4©; the DC side inductance 5© and the storage system 6©. In order to derive a statespace model, controlled variables are shown in the Figure 1a where:

- X1 stands for output voltage, or grid voltage while it is connected to the grid;- X2 represents the inverter current;- X3 is the common DC capacitor voltage;- X4 is the DC storage current;- VB is the storage system voltage, to derive the model for simplicity, and it is considered constant;- s1 and s2 are complementary gate signals to derive the IGBT bridge, and those are either zero

or one;- DBuck and DBoost are not complementary gate signals; to derive the chopper leg, those are either

zero or one.

Energies 2016, 9, 233 3 of 14Energies 2016, 9, 233 3 of 14

Figure 1. Bidirectional converter (a) schematic and (b) prototype.

Controlled switches gate signals, according to [27], are modeled by continuous functions and the corresponding variable will be shown by ŝ1, ŝ2, BuckD and BoostD . Thus, to derive the equations and the model, instead of discontinuous zero and one function, the continuous function will be used.

The inverter side model can be driven by Kirchhoff's Voltage Law (KVL) in loop I and Kirchhoff's Current Law (KCL) on the DC bus capacitor current Equations (1) and (2).

0ˆˆ 3231 XsXsVV Lacs (1)

0ˆˆ 2221 XsXsic (2)

which, using state space parameters and rewriting the equations, lead to

323121 ˆˆ XsXsXSLX ac (3)

22213 ˆˆ XsXsXSCT (4)

S stands for Laplace transform parameter, and the complementary properties of ŝ1 and ŝ2 necessitates that: ŝ1 + ŝ2 = 1. Using this property, the inverter model can be simplified.

The DC chopper leg during charge mode works as Buck converter and during the discharge mode as Boost one. Thus, in the model, both modes of operations have been considered, and, for each operation, mode specific equations have to be used. Using the same notation, we can write the KVL of loop II for chopper leg Equation (5).

0 swLdcB VVV (5)

where Vsw is defined as switch side equivalent voltage for either Boost and Buck conditions. Replacing the Boost and Buck voltage, Equations (5) and (6) can be rewritten for two different operation modes as reported in Equation (7). Using Laplace transform for inductance voltage, Equation (7) leads to

Figure 1. Bidirectional converter (a) schematic and (b) prototype.

Controlled switches gate signals, according to [27], are modeled by continuous functions and thecorresponding variable will be shown by s1, s2, DBuck and DBoost. Thus, to derive the equations andthe model, instead of discontinuous zero and one function, the continuous function will be used.

The inverter side model can be driven by Kirchhoff’s Voltage Law (KVL) in loop I and Kirchhoff’sCurrent Law (KCL) on the DC bus capacitor current Equations (1) and (2).

Vs `VLac ´ s1X3 ` s2X3 “ 0 (1)

ic ´ s1X2 ` s2X2 “ 0 (2)

which, using state space parameters and rewriting the equations, lead to

X1 ` SLacX2 “ s1X3 ´ s2X3 (3)

SCTX3 “ s1X2 ´ s2X2 (4)

S stands for Laplace transform parameter, and the complementary properties of s1 and s2

necessitates that: s1 + s2 = 1. Using this property, the inverter model can be simplified.The DC chopper leg during charge mode works as Buck converter and during the discharge

mode as Boost one. Thus, in the model, both modes of operations have been considered, and, for eachoperation, mode specific equations have to be used. Using the same notation, we can write the KVL ofloop II for chopper leg Equation (5).

VB ´VLdc ´Vsw “ 0 (5)

Energies 2016, 9, 233 4 of 14

where Vsw is defined as switch side equivalent voltage for either Boost and Buck conditions. Replacingthe Boost and Buck voltage, Equations (5) and (6) can be rewritten for two different operation modesas reported in Equation (7). Using Laplace transform for inductance voltage, Equation (7) leads toEquation (8) and applying Boost and Buck converter equation, the current balance can be written asreported in Equation (9), where iC is the current through the DC link capacitor. Therefore, voltage andcurrent equations for Boost and Buck converters result in:

Boost BuckVsw “ p1´ DBoostqX3 Vsw “ DBuckX3 (6)

VB ´VLdc ´ p1´ DBoostqX3 “ 0 VB ´VLdc ´ DBuckX3 “ 0 (7)VB ´ SLDCX4 ´ p1´ DBoostqX3 “ 0 VB ´ SLDCX4 ´ DBuckX3 “ 0 (8)

iC “ SCTX3 “`

1´ DBoost˘

iDC SCTX3 “ DBuck iDC (9)

By Equations (3), (4), (8) and (9), the state space model of the bidirectional converter can be driven.Due to multiplication of time varying switching functions and duty cycles to state variables, it can beconsidered as Linear Time Varying (LTV) system, and the model can be solved by the proper iterationmethod. There are two main operation modes for these converters, charging the storage only if it isconnected to the grid, and, in this case, the model output is X2 (the inverter current). Dischargingthe storage supplying loads by using energy stored in the storage where the X1 (the inverter outputvoltage) is considered as model output. In each case, all independent state variables are computedby duty cycles, DBuck or DBoost and continuous switching functions s1 and s2. The model is shown tounderstand the converter behavior better, and some equations are used for controller design purpose.The complete model analysis will be addressed in another publication.

3. Control Method

The converter control, as presented in Figure 1a, has two separate parts—chopper leg controland inverter control. These two parts practically are linked together through DC bus capacitor.Two controllers are presented here separately. Chopper and inverter control are based on amodified standard method introducing an MPCC, and the designed control logics are describedin subsections that follow. For controller design, some equation from a derived model is adopted fordigital implementation.

3.1. Chopper Control

Chopper leg can be managed either to charge the storage or to discharge the storage in order tosupply the load.

3.1.1. Charging the Storage

As it is seen from Figure 1a, the battery management leg has two IGBTs, where the midpoint isconnected to the batteries’ positive terminal through an inductance. Considering antiparallel diodesof the IGBTs by blocking the low-side switch and controlling the high-side switch, it becomes Buckconverter, which can charge the batteries using voltage through DC bus capacitor. Standard batterycharging is applied in order to charge the batteries, which is called the Constant Current ConstantVoltage (CCCV) method. This technique of battery charging has two modes of charging that dependon batteries’ state of charges—constant current and constant voltage [28]. It charges the batteries untilnear nominal voltage by inserting constant current. Then, it uses constant voltage, charges the batterieskeeping the voltage constant, meanwhile the current goes to zero slowly [28].

For the constant current mode by using the basic principle of Buck converter operation [29],considering the equation (8-Buck) part and using Euler first order approximation instead of Laplacetransfer function for the derivative, the voltage balance on inductance for one switching period can beachived by rewriting (8-Buck). Therefore, the duty cycle can be evaluated by the Equation (10).

Energies 2016, 9, 233 5 of 14

DBuck “

ˆ

´∆iLDC ˆ LDC

Ts`VB

˙

VDC“

ˆ

´piLDCpkq ´ iLDCpk´ 1qq ˆ LDC

Ts`VB

˙

VDC(10)

where iLDC is the DC side inductance current in ampere; k is the number related to time step; LDCis the DC side inductance value in Henry; VB is the instantaneous overall battery voltage in volt;VDC is the inverter DC bus voltage; Ts is the controller sampling time in second, and DBuck is thecalculated duty cycle in p.u. Considering the predictive controller, ∆iLDC has to be replaced withthe difference between reference current (in our case, the batteries’ nominal current equal to 3A) andmeasured current. In so doing, Equation (10) changes to Equation (11). The generated duty cycle runsthe high-side switch of the chopper leg:

DBuck “

¨

˝´

´

ire f ´ iLDCpkq¯

ˆ LDC

Ts`VB

˛

‚

VDC(11)

For constant voltage charging control, since the batteries’ voltage is near nominal, the duty cyclecan be considered as Buck converter voltage Equation (6); however, in order to have better control onthe absorbed current, a cascade controller is used. The controller schema is shown in Figure 2.

Energies 2016, 9, 233 5 of 14

where iLDC is the DC side inductance current in ampere; k is the number related to time step; LDC is the DC side inductance value in Henry; VB is the instantaneous overall battery voltage in volt; VDC is the inverter DC bus voltage; Ts is the controller sampling time in second, and DBuck is the calculated duty cycle in p.u. Considering the predictive controller, ΔiLDC has to be replaced with the difference between reference current (in our case, the batteries’ nominal current equal to 3A) and measured current. In so doing, Equation (10) changes to Equation (11). The generated duty cycle runs the high-side switch of the chopper leg:

DC

Bs

DCLDCref

Buck V

VT

Lkii

D

)(

(11)

For constant voltage charging control, since the batteries’ voltage is near nominal, the duty cycle can be considered as Buck converter voltage Equation (6); however, in order to have better control on the absorbed current, a cascade controller is used. The controller schema is shown in Figure 2.

Figure 2. Constant voltage chopper charging controller—PI-MPCC, chopper.

The difference between nominal voltage value and measured voltage is fed to a PI controller, and the output is considered as the current reference, which is used as reference to the constant current controller of Equation (10). In order to have control over the absorbed current, PI (VB) output is limited with upper and lower limits. It is considered as batteries’ charging current, so the max is cut to the batteries nominal charging current. Switching and sampling time of charging mode for DC chopper leg IGBTs are the same as the inverter ones (20 kHz).

3.1.2. Discharging Mode

In the discharge mode, the high-side switch of the DC chopper leg is blocked, and by running the lower switch and means of the antiparallel diode, it works as a Boost converter. During discharging mode, the batteries’ voltage is boosted to a higher value and keeps DC bus voltage constant for inverter functioning to supply the load. During discharging mode, in order to decrease losses and keep the converter at Discontinuous Current operation Mode (DCM), the switching frequency is considered much lower than the inverter one. Control approach is based on cascade PI controller with outer voltage control and inner current controller. The schema is shown in Figure 3.

Figure 3. Chopper Discharging controller.

The second PI output is considered as the duty cycle to run the corresponding switch. First PI, PI (VDC) output is the reference DC current for battery discharging mode. During discharge mode, the battery current can be high and the limit is set corresponding to DC inductance rated value to avoid inductance saturation. PI (IDC) output is the duty cycle for the PWM module, so it is limited to stay below one.

Figure 2. Constant voltage chopper charging controller—PI-MPCC, chopper.

The difference between nominal voltage value and measured voltage is fed to a PI controller, andthe output is considered as the current reference, which is used as reference to the constant currentcontroller of Equation (10). In order to have control over the absorbed current, PI (VB) output is limitedwith upper and lower limits. It is considered as batteries’ charging current, so the max is cut to thebatteries nominal charging current. Switching and sampling time of charging mode for DC chopperleg IGBTs are the same as the inverter ones (20 kHz).

3.1.2. Discharging Mode

In the discharge mode, the high-side switch of the DC chopper leg is blocked, and by running thelower switch and means of the antiparallel diode, it works as a Boost converter. During dischargingmode, the batteries’ voltage is boosted to a higher value and keeps DC bus voltage constant for inverterfunctioning to supply the load. During discharging mode, in order to decrease losses and keep theconverter at Discontinuous Current operation Mode (DCM), the switching frequency is consideredmuch lower than the inverter one. Control approach is based on cascade PI controller with outervoltage control and inner current controller. The schema is shown in Figure 3.

Energies 2016, 9, 233 6 of 14

Energies 2016, 9, 233 5 of 14

where iLDC is the DC side inductance current in ampere; k is the number related to time step; LDC is the DC side inductance value in Henry; VB is the instantaneous overall battery voltage in volt; VDC is the inverter DC bus voltage; Ts is the controller sampling time in second, and DBuck is the calculated duty cycle in p.u. Considering the predictive controller, ΔiLDC has to be replaced with the difference between reference current (in our case, the batteries’ nominal current equal to 3A) and measured current. In so doing, Equation (10) changes to Equation (11). The generated duty cycle runs the high-side switch of the chopper leg:

DC

Bs

DCLDCref

Buck V

VT

Lkii

D

)(

(11)

For constant voltage charging control, since the batteries’ voltage is near nominal, the duty cycle can be considered as Buck converter voltage Equation (6); however, in order to have better control on the absorbed current, a cascade controller is used. The controller schema is shown in Figure 2.

Figure 2. Constant voltage chopper charging controller—PI-MPCC, chopper.

The difference between nominal voltage value and measured voltage is fed to a PI controller, and the output is considered as the current reference, which is used as reference to the constant current controller of Equation (10). In order to have control over the absorbed current, PI (VB) output is limited with upper and lower limits. It is considered as batteries’ charging current, so the max is cut to the batteries nominal charging current. Switching and sampling time of charging mode for DC chopper leg IGBTs are the same as the inverter ones (20 kHz).

3.1.2. Discharging Mode

In the discharge mode, the high-side switch of the DC chopper leg is blocked, and by running the lower switch and means of the antiparallel diode, it works as a Boost converter. During discharging mode, the batteries’ voltage is boosted to a higher value and keeps DC bus voltage constant for inverter functioning to supply the load. During discharging mode, in order to decrease losses and keep the converter at Discontinuous Current operation Mode (DCM), the switching frequency is considered much lower than the inverter one. Control approach is based on cascade PI controller with outer voltage control and inner current controller. The schema is shown in Figure 3.

Figure 3. Chopper Discharging controller.

The second PI output is considered as the duty cycle to run the corresponding switch. First PI, PI (VDC) output is the reference DC current for battery discharging mode. During discharge mode, the battery current can be high and the limit is set corresponding to DC inductance rated value to avoid inductance saturation. PI (IDC) output is the duty cycle for the PWM module, so it is limited to stay below one.

Figure 3. Chopper Discharging controller.

The second PI output is considered as the duty cycle to run the corresponding switch. First PI,PI (VDC) output is the reference DC current for battery discharging mode. During discharge mode,the battery current can be high and the limit is set corresponding to DC inductance rated value toavoid inductance saturation. PI (IDC) output is the duty cycle for the PWM module, so it is limited tostay below one.

3.2. Inverter Control Method

Inverter control also has two states. When the inverter is connected in parallel to the grid,the inverter terminal voltage is imposed by the grid so the inverter needs to work as the current source.Instead, when the inverter is disconnected from the grid, it should be able to keep its terminal voltagewith constant frequency and magnitude under wide load variation. Both states will be discussed belowin detail. Indeed, both controllers use a current controller as the latest controller in their loop and thispaper used a Model Predictive Current Controller (MPCC) as current controller. Therefore, the MPCCis described before presenting two operational modes controllers. The focus of this study is to managethe inverter properly, so it will be able to follow any current and voltage reference.

3.2.1. Model Predictive Current Controller (MPCC)

The implemented digital MPCC is based on an inverter model, KVL of loop I in Figure 1b.The MPCC is designed based on Equation (1). The Vs and X3 are network voltage and DC bus voltagesrespectively. Inside the controller, these two variables are real measurments. The VLac and s1 ands2 are unknown so, to calculate switching functions, s1 and s2, it is necessary to estimate VLac. Thetraditional method to estimate voltage drop on the inductance is using Euler first order approximationfor derivative. Therefore, it can be written in the discrete domain as:

vL “∆iL ˆ Lac

Ts“piLpkq ´ iLpk´ 1qq ˆ Lac

Ts(12)

where k is the iteration order, iL(k) and iL(k-1) are inductance currents at corresponding iterations. Lac isthe coupling inductance value in Henry and Ts is the iteration step (in this work, it is equal to 50 µsboth in simulation and experiment). All the variables in Equation (12) are presented in discrete timedomain. In order to control the current in the coupling inductance, it is assumed that the inverter worksin a stable condition, and it follows the reference current properly. With this assumption, it is possibleto consider the reference current as a prediction for the next iteration step of the inverter current andupdate Equation (12). Therefore, the current deviation on the inductance can be represented as thedifference of reference current and measured current.

∆iL “ iLpkq ´ iLpk´ 1q “ ire f ´ iLpkq (13)

where the ire f is the instantaneous reference current. With this estimation, the voltage drop on theinductance can be predicted by adopting Equation (14) instead of Equation (12). Equation (14) can bereplaced in Equation (1) for digital implementation. Rewriting Equation (1) using Equation (14) andusing a complimentary property of the inverter continues switching functions, and the continuousinverter switching function can be computed by Equation (15):

Energies 2016, 9, 233 7 of 14

vL “

´

ire f ´ iLpkq¯

ˆ Lac

Ts(14)

s1 “pvs ` vDC ` vLq

2ˆ vDC(15)

The switching function evaluated by Equation (15) s1 and s2 = 1 ´ s1 are fed to the PWM signalgenerator to run the IGBT drivers of inverter legs. This MPCC has a fast and reliable response of themodel predictive controller, and it also has constant switching frequency due to the PWM techniqueintegration into the control loop. This controller is used in both grid connected and island operationmode controllers.

3.2.2. Grid Connected Mode

During grid connected mode, the full bridge inverter should manage to keep capacitor DC voltageconstant to charge the storage, if it is necessary. Depending on the application, the inverter can also beresponsible for several tasks like reactive power compensation, harmonic elimination, load sharing,and so on. All control methods, in order to manage reactive current, harmonics and voltage issues,calculate reference current or voltage that the inverter should follow. During the grid connected mode,the device has to work as current source and needs to properly follow a reference current. Figure 4shows the grid connected operation mode controller block diagram, which is the described MPCC fedto the PWM module of microcontroller.

Energies 2016, 9, 233 7 of 14

voltage issues, calculate reference current or voltage that the inverter should follow. During the grid connected mode, the device has to work as current source and needs to properly follow a reference current. Figure 4 shows the grid connected operation mode controller block diagram, which is the described MPCC fed to the PWM module of microcontroller.

Figure 4. Grid connected mode controller block diagram.

In this mode, DC bus voltage has been kept constant by using a PI controller. The difference between reference value and measured voltage fed to a PI controller as the error. The output added up to the inverter reference current.

3.2.3. Island Mode

During Island operation mode, the inverter terminal voltage should be controlled as well. Therefore, the inverter voltage PI controller is added in the front end of the described MPCC to design the proposed PI-MPCC. Since inverter current measurement is used for the MPCC inverter current controller, the same measure can therefore be used in the island voltage controller. A control method, based on cascade controller with load current feedback, is implemented in order to avoid voltage drop under load variation for island operation mode [30]. The implemented controller is illustrated in Figure 5.

Figure 5. Island mode controller block diagram—PI-MPCC, inverter.

An ideal sinusoidal waveform with nominal frequency and peak value is considered as the voltage reference, it is compared instantaneously with inverter AC terminal voltage, and the error is fed to a PI controller. The PI (Vinv) output is considered as the control current, and iControl is added to the current reference for the inverter current controller. The limit is set according to AC side inductance current ratio. iControl is added to the load current feedback and the estimated output filter capacitor current, which uses the reference voltage and basic capacitor equation in order to estimate capacitor current. The sum is considered as the reference current. The generated reference current is fed to the inverter MPCC that is explained in subsection 3.2.1.

4. Results

In order to validate the control method, the derived model and the introduced control method has been implemented and some computer simulations have been developed and compared with the experimental results. To observe inverter stability and performance, the system has been tested under different scenarios.

4.1. System Description

The model has been simulated by using MATLAB SimPowerSystems toolbox. Ideal IGBTs with antiparallel diodes are used to build full bridge inverters. The MPCC explained in Section 3 is used to run the inverter. The inverter rated power is chosen as 3 kW for the experimental setup. Simulated

Figure 4. Grid connected mode controller block diagram.

In this mode, DC bus voltage has been kept constant by using a PI controller. The differencebetween reference value and measured voltage fed to a PI controller as the error. The output added upto the inverter reference current.

3.2.3. Island Mode

During Island operation mode, the inverter terminal voltage should be controlled as well.Therefore, the inverter voltage PI controller is added in the front end of the described MPCC todesign the proposed PI-MPCC. Since inverter current measurement is used for the MPCC invertercurrent controller, the same measure can therefore be used in the island voltage controller. A controlmethod, based on cascade controller with load current feedback, is implemented in order to avoidvoltage drop under load variation for island operation mode [30]. The implemented controller isillustrated in Figure 5.

Energies 2016, 9, 233 7 of 14

voltage issues, calculate reference current or voltage that the inverter should follow. During the grid connected mode, the device has to work as current source and needs to properly follow a reference current. Figure 4 shows the grid connected operation mode controller block diagram, which is the described MPCC fed to the PWM module of microcontroller.

Figure 4. Grid connected mode controller block diagram.

In this mode, DC bus voltage has been kept constant by using a PI controller. The difference between reference value and measured voltage fed to a PI controller as the error. The output added up to the inverter reference current.

3.2.3. Island Mode

During Island operation mode, the inverter terminal voltage should be controlled as well. Therefore, the inverter voltage PI controller is added in the front end of the described MPCC to design the proposed PI-MPCC. Since inverter current measurement is used for the MPCC inverter current controller, the same measure can therefore be used in the island voltage controller. A control method, based on cascade controller with load current feedback, is implemented in order to avoid voltage drop under load variation for island operation mode [30]. The implemented controller is illustrated in Figure 5.

Figure 5. Island mode controller block diagram—PI-MPCC, inverter.

An ideal sinusoidal waveform with nominal frequency and peak value is considered as the voltage reference, it is compared instantaneously with inverter AC terminal voltage, and the error is fed to a PI controller. The PI (Vinv) output is considered as the control current, and iControl is added to the current reference for the inverter current controller. The limit is set according to AC side inductance current ratio. iControl is added to the load current feedback and the estimated output filter capacitor current, which uses the reference voltage and basic capacitor equation in order to estimate capacitor current. The sum is considered as the reference current. The generated reference current is fed to the inverter MPCC that is explained in subsection 3.2.1.

4. Results

In order to validate the control method, the derived model and the introduced control method has been implemented and some computer simulations have been developed and compared with the experimental results. To observe inverter stability and performance, the system has been tested under different scenarios.

4.1. System Description

The model has been simulated by using MATLAB SimPowerSystems toolbox. Ideal IGBTs with antiparallel diodes are used to build full bridge inverters. The MPCC explained in Section 3 is used to run the inverter. The inverter rated power is chosen as 3 kW for the experimental setup. Simulated

Figure 5. Island mode controller block diagram—PI-MPCC, inverter.

Energies 2016, 9, 233 8 of 14

An ideal sinusoidal waveform with nominal frequency and peak value is considered as the voltagereference, it is compared instantaneously with inverter AC terminal voltage, and the error is fed to a PIcontroller. The PI (Vinv) output is considered as the control current, and iControl is added to the currentreference for the inverter current controller. The limit is set according to AC side inductance currentratio. iControl is added to the load current feedback and the estimated output filter capacitor current,which uses the reference voltage and basic capacitor equation in order to estimate capacitor current.The sum is considered as the reference current. The generated reference current is fed to the inverterMPCC that is explained in Subsection 3.2.1.

4. Results

In order to validate the control method, the derived model and the introduced control methodhas been implemented and some computer simulations have been developed and compared with theexperimental results. To observe inverter stability and performance, the system has been tested underdifferent scenarios.

4.1. System Description

The model has been simulated by using MATLAB SimPowerSystems toolbox. Ideal IGBTs withantiparallel diodes are used to build full bridge inverters. The MPCC explained in Section 3 is used torun the inverter. The inverter rated power is chosen as 3 kW for the experimental setup. Simulatedand experimental prototype parameters are listed in Table 1. In order to control the gate signal ofIGBTs and run the system in the experimental prototype, Texas Instruments (TI) TMS320F28335 basedDSP controller board is used for gate signals, and it is fed through the IGBT driver board. UsingTMS320F28335 PWM output signals and also implementing the driver board to run IGBTs insertsproper dead time in order to avoid short circuit on DC bus. The driver board by means of a transformer,guarantees metallic isolation between the power circuit and control board.

Table 1. Single phase bidirectional converter parameters.

Grid Nominal Voltage 230 V rms

Nominal frequency 50 HzStorage voltage 80 VDC bus voltage 450 V

DC bus capacitor C 3 parallel 6800 µFDC side switching inductance L 1 mHac side switching inductance L 1 mH

Inverter output filter 10 µFMax power switches (IGBT) 195 A

Batteries 12 V, 12 Ah, Sealed Lead-AcidSampling time 10 µs

The controllers’ set up is reported in Table 2. Controller nomenclatures are as shown inFigures 2, 3 and 5. The saturation limit of controllers are also reported in Table 2.

Table 2. Bidirectional converter controller set up.

kp kiOutput Saturation

Item Max Min

PI (VB) 0.5 35 IDC_re f 3 A 0 API (VDC) 3 25 IDC_re f 45 A 0 API (IDC) 12 50 Duty cycle 0.95 0PI (Vinv) 0.74 3000 IControl 35 A ´35 A

Energies 2016, 9, 233 9 of 14

The performance of the system has been tested under several scenarios both in simulations andexperiments, and results are presented together—first, the grid connected mode charging the batteries,and second, island mode, using energy stored in batteries to supply the load. In the grid connectedmode, the inverter works as a current source. In Island operation mode, it works as an ideal voltagesource using stored energy inside the batteries.

4.2. Scenario I: Grid Connected, No Load

In this configuration, the device is connected in parallel to the grid and it can work as a parallelconnected power quality conditioner giving services to the grid and load. Here, the device performanceto work in grid connected mode, with no load, is presented. The device just tries to keep the inverterDC bus voltage constant, compensating its losses and supporting the chopper leg to charge the batteries.The chopper leg charges the batteries with constant 3 A at around 80 V so, the inverter absorbs about1.6 A ac current from network. Simulation results for the grid voltage and grid side current are shownin Figure 6a.Energies 2016, 9, 233 9 of 14

Figure 6. Grid connected inverter response—no load, (a) Simulation; (b) Experiment.

Figure 6b shows the experimental results with no load. The system just manages to charge the batteries. The system absorbs current around 1.7 A rms from the networks, which is equal to batteries nominal charging current at DC 80 V voltage plus the current required to compensate system losses. There is a 0.1 A difference between simulation and experimental results, which could be because of acquisition noises or measurement tool adjustment. Reported experimental data are recorded with differential probes and THD of the voltage is 1.59%, and for the current is equal to 11.80%. The THD analysis for simulation study is much better because, for simulation, the grid voltage is considered as the ideal voltage source (THD = 0), and the control method is also not affected by acquisition noises so the THD of the current from simulation is equal to 4.81%.

4.3. Scenario II: Grid Connected, 1 kW Resistive Load

As the second illustration, the device is connected to a linear 1 kW resistive load (five 200 W incandescent lamps in experimental test). The grid side current, inverter current and grid voltage from simulation are shown in Figure 7a. At this operation mode, the device can give several services to the load and grid [4]. In this mode, the inverter keeps the DC bus capacitor voltage constant and charges the batteries. In simulation, the grid voltage is considered 215 V to compare with experimental results (during the experimental tests, the grid voltage was 215 V). Thus, the grid absorbs about 5.8 A current. It is the summation of load current and the inverter current. THD results of the simulation are 4.81% and 2.17% for inverter current and grid current, respectively.

Figure 7. Grid connected inverter response—1 kW resistive load, (a) Simulation; (b) Experiment.

Figure 7b shows the experiment’s results. The grid supplies are around 5.7 A rms current, which is the summation of load current and also batteries’ charging current. In this case, experimental results’ THD analysis is 1.48%, 24.52% and 3.01% for grid voltage, inverter current and grid current,

Figure 6. Grid connected inverter response—no load, (a) Simulation; (b) Experiment.

Figure 6b shows the experimental results with no load. The system just manages to charge thebatteries. The system absorbs current around 1.7 A rms from the networks, which is equal to batteriesnominal charging current at DC 80 V voltage plus the current required to compensate system losses.There is a 0.1 A difference between simulation and experimental results, which could be because ofacquisition noises or measurement tool adjustment. Reported experimental data are recorded withdifferential probes and THD of the voltage is 1.59%, and for the current is equal to 11.80%. The THDanalysis for simulation study is much better because, for simulation, the grid voltage is considered asthe ideal voltage source (THD = 0), and the control method is also not affected by acquisition noises sothe THD of the current from simulation is equal to 4.81%.

4.3. Scenario II: Grid Connected, 1 kW Resistive Load

As the second illustration, the device is connected to a linear 1 kW resistive load (five 200 Wincandescent lamps in experimental test). The grid side current, inverter current and grid voltage fromsimulation are shown in Figure 7a. At this operation mode, the device can give several services to theload and grid [4]. In this mode, the inverter keeps the DC bus capacitor voltage constant and chargesthe batteries. In simulation, the grid voltage is considered 215 V to compare with experimental results(during the experimental tests, the grid voltage was 215 V). Thus, the grid absorbs about 5.8 A current.It is the summation of load current and the inverter current. THD results of the simulation are 4.81%and 2.17% for inverter current and grid current, respectively.

Energies 2016, 9, 233 10 of 14

Energies 2016, 9, 233 9 of 14

Figure 6. Grid connected inverter response—no load, (a) Simulation; (b) Experiment.

Figure 6b shows the experimental results with no load. The system just manages to charge the batteries. The system absorbs current around 1.7 A rms from the networks, which is equal to batteries nominal charging current at DC 80 V voltage plus the current required to compensate system losses. There is a 0.1 A difference between simulation and experimental results, which could be because of acquisition noises or measurement tool adjustment. Reported experimental data are recorded with differential probes and THD of the voltage is 1.59%, and for the current is equal to 11.80%. The THD analysis for simulation study is much better because, for simulation, the grid voltage is considered as the ideal voltage source (THD = 0), and the control method is also not affected by acquisition noises so the THD of the current from simulation is equal to 4.81%.

4.3. Scenario II: Grid Connected, 1 kW Resistive Load

As the second illustration, the device is connected to a linear 1 kW resistive load (five 200 W incandescent lamps in experimental test). The grid side current, inverter current and grid voltage from simulation are shown in Figure 7a. At this operation mode, the device can give several services to the load and grid [4]. In this mode, the inverter keeps the DC bus capacitor voltage constant and charges the batteries. In simulation, the grid voltage is considered 215 V to compare with experimental results (during the experimental tests, the grid voltage was 215 V). Thus, the grid absorbs about 5.8 A current. It is the summation of load current and the inverter current. THD results of the simulation are 4.81% and 2.17% for inverter current and grid current, respectively.

Figure 7. Grid connected inverter response—1 kW resistive load, (a) Simulation; (b) Experiment.

Figure 7b shows the experiment’s results. The grid supplies are around 5.7 A rms current, which is the summation of load current and also batteries’ charging current. In this case, experimental results’ THD analysis is 1.48%, 24.52% and 3.01% for grid voltage, inverter current and grid current,

Figure 7. Grid connected inverter response—1 kW resistive load, (a) Simulation; (b) Experiment.

Figure 7b shows the experiment’s results. The grid supplies are around 5.7 A rms current, whichis the summation of load current and also batteries’ charging current. In this case, experimentalresults’ THD analysis is 1.48%, 24.52% and 3.01% for grid voltage, inverter current and grid current,respectively. Although the inverter current has a high THD factor, on the grid side, the current andvoltage THDs are within standard range.

The difference between simulation and experimental results is again due to acquisition noises,which are apparent in grid voltage and inverter current at Figure 7b and also measurement tools.It can especially be noticed from Figure 7b that the grid current measurement clamp bandwidth is notwide enough to catch switching ripples.

4.4. Scenario III: Island Mode, No Load

The transition from online operation mode to island operation mode is out of the scope of thispaper. It is addressed in previous publications [31,32], so it is assumed that the system is working withno load at island operation mode at the steady state. The controllers set up are as reported in Table 2.Simulation results for the voltage at the inverter terminals are shown in Figure 8a. In this operationmode, the inverter uses energy stored in batteries to work as an ideal voltage source. Inverter voltageTHD for simulation is 0.46%.

Energies 2016, 9, 233 10 of 14

respectively. Although the inverter current has a high THD factor, on the grid side, the current and voltage THDs are within standard range.

The difference between simulation and experimental results is again due to acquisition noises, which are apparent in grid voltage and inverter current at Figure 7b and also measurement tools. It can especially be noticed from Figure 7b that the grid current measurement clamp bandwidth is not wide enough to catch switching ripples.

4.4. Scenario III: Island Mode, No Load

The transition from online operation mode to island operation mode is out of the scope of this paper. It is addressed in previous publications [31,32], so it is assumed that the system is working with no load at island operation mode at the steady state. The controllers set up are as reported in Table 2. Simulation results for the voltage at the inverter terminals are shown in Figure 8a. In this operation mode, the inverter uses energy stored in batteries to work as an ideal voltage source. Inverter voltage THD for simulation is 0.46%.

Figure 8. Island operation mode—no load, (a) Simulation; (b) Experiment.

The experimental testbed is also tested in Island operation mode. With no load, the voltage profile is shown in Figure 8b. The inverter keeps output voltage constant with 230 V rms at 50 Hz. THD analysis results for output voltage are equal to 2.29%. Simulation and experimental results are well matched, although, again, the experimental THD is higher than the simulation one due to the reasons that are already addressed.

4.5. Scenario IV: Island Mode, 1 kW Resistive Load

Figure 9b shows the simulated system response in island operation mode under the same 1 kW load. It shows the inverter output voltage and current in steady state condition. The inverter supplies 4.3 A current. The THD analysis of simulation results for the inverter voltage and current are 0.43% and 3.81%, respectively.

The experimental setup is also run in Island mode, supplying 1 kW resistive load (five 200 W incandescent lamps). The results are shown in Figure 9b. The inverter keeps load voltage at a constant 230 V rms, supplying around 4.3 A rms current (same as simulation result). There is phase displacement between load current and voltage because the inverter output passive filter needs reactive power. THD analysis results are 2.28% for inverter voltage and 6.52% for inverter current. It should be noted that voltage differential probes and current clamps are used to record the experimental data. The inverter current is recorded by a DC clamp with low bandwidth (20 kHz) so, the inverter current switching ripples could not be recorded. This is the reason why in Figure 9 the experimental results outperform the simulation ones

Figure 8. Island operation mode—no load, (a) Simulation; (b) Experiment.

The experimental testbed is also tested in Island operation mode. With no load, the voltage profileis shown in Figure 8b. The inverter keeps output voltage constant with 230 V rms at 50 Hz. THDanalysis results for output voltage are equal to 2.29%. Simulation and experimental results are wellmatched, although, again, the experimental THD is higher than the simulation one due to the reasonsthat are already addressed.

Energies 2016, 9, 233 11 of 14

4.5. Scenario IV: Island Mode, 1 kW Resistive Load

Figure 9b shows the simulated system response in island operation mode under the same 1 kWload. It shows the inverter output voltage and current in steady state condition. The inverter supplies4.3 A current. The THD analysis of simulation results for the inverter voltage and current are 0.43%and 3.81%, respectively.Energies 2016, 9, 233 11 of 14

Figure 9. Island operation mode, inverter response—1 kW resistive load, (a) Simulation; (b) Experiment.

Figure 10 shows the transient behavior on DC bus voltage when the load changes from zero to 1 kW. In Figure 10a, the system was working in island mode with no load until t = 3.7 s. At t = 3.7 s, the 1 kW load is connected as the load. As can be seen, there are the voltage drops by a few volts with DC bus voltage. The system is able to recover this voltage drop within less than half a second. This transient response can be much faster, however, the DC battery current absorption is limited inside the controller to avoid any battery damage.

Figure 10. Island operation mode, DC bus transient response—1 kW resistive load, (a) Simulation; (b) Experiment.

Figure 10b illustrates the experimental setup DC bus voltage response once the 1 kW load is connected to the system. It shows that the system’s response is fast enough to keep DC bus voltage at a set point. The voltage oscillation during the transient compared to simulation results is higher because the simulation is considered an ideal system where there is no acquisition problem and noises. However, in practice, as it can be noticed from Figure 10b, the DC bus has much more noise and oscillation with respect to the simulation one. It is possible to notice the difference between simulation and experimental results in Figure 10b, because, in this case, the voltage probe has higher 30–40 MHz bandwidth. The results are meant to show steady state performance, and the transient behavior was not focused on in this paper.

5. Conclusions

Simple MPCC, utilizing advantages of the conventional PI controller, has been designed for a single phase bidirectional converter. With the proposed PI-MPCC controller variable, the switching function of the original MPC is improved. Moreover, the optimization cost function can be easily adopted to the controller based on the application. The converter is the basic component of most renewable energy types and a major part of the micro grid and smart grid concept. The control method has been tested both in MATLAB simulation environment and with a 3 kW experimental prototype. Results show that the inverter has very good transient and steady state performance. The

Figure 9. Island operation mode, inverter response—1 kW resistive load, (a) Simulation;(b) Experiment.

The experimental setup is also run in Island mode, supplying 1 kW resistive load (five 200 Wincandescent lamps). The results are shown in Figure 9b. The inverter keeps load voltage at a constant230 V rms, supplying around 4.3 A rms current (same as simulation result). There is phase displacementbetween load current and voltage because the inverter output passive filter needs reactive power. THDanalysis results are 2.28% for inverter voltage and 6.52% for inverter current. It should be noted thatvoltage differential probes and current clamps are used to record the experimental data. The invertercurrent is recorded by a DC clamp with low bandwidth (20 kHz) so, the inverter current switchingripples could not be recorded. This is the reason why in Figure 9 the experimental results outperformthe simulation ones

Figure 10 shows the transient behavior on DC bus voltage when the load changes from zero to1 kW. In Figure 10a, the system was working in island mode with no load until t = 3.7 s. At t = 3.7 s,the 1 kW load is connected as the load. As can be seen, there are the voltage drops by a few volts withDC bus voltage. The system is able to recover this voltage drop within less than half a second. Thistransient response can be much faster, however, the DC battery current absorption is limited inside thecontroller to avoid any battery damage.

Energies 2016, 9, 233 11 of 14

Figure 9. Island operation mode, inverter response—1 kW resistive load, (a) Simulation; (b) Experiment.

Figure 10 shows the transient behavior on DC bus voltage when the load changes from zero to 1 kW. In Figure 10a, the system was working in island mode with no load until t = 3.7 s. At t = 3.7 s, the 1 kW load is connected as the load. As can be seen, there are the voltage drops by a few volts with DC bus voltage. The system is able to recover this voltage drop within less than half a second. This transient response can be much faster, however, the DC battery current absorption is limited inside the controller to avoid any battery damage.

Figure 10. Island operation mode, DC bus transient response—1 kW resistive load, (a) Simulation; (b) Experiment.

Figure 10b illustrates the experimental setup DC bus voltage response once the 1 kW load is connected to the system. It shows that the system’s response is fast enough to keep DC bus voltage at a set point. The voltage oscillation during the transient compared to simulation results is higher because the simulation is considered an ideal system where there is no acquisition problem and noises. However, in practice, as it can be noticed from Figure 10b, the DC bus has much more noise and oscillation with respect to the simulation one. It is possible to notice the difference between simulation and experimental results in Figure 10b, because, in this case, the voltage probe has higher 30–40 MHz bandwidth. The results are meant to show steady state performance, and the transient behavior was not focused on in this paper.

5. Conclusions

Simple MPCC, utilizing advantages of the conventional PI controller, has been designed for a single phase bidirectional converter. With the proposed PI-MPCC controller variable, the switching function of the original MPC is improved. Moreover, the optimization cost function can be easily adopted to the controller based on the application. The converter is the basic component of most renewable energy types and a major part of the micro grid and smart grid concept. The control method has been tested both in MATLAB simulation environment and with a 3 kW experimental prototype. Results show that the inverter has very good transient and steady state performance. The

Figure 10. Island operation mode, DC bus transient response—1 kW resistive load, (a) Simulation;(b) Experiment.

Energies 2016, 9, 233 12 of 14

Figure 10b illustrates the experimental setup DC bus voltage response once the 1 kW load isconnected to the system. It shows that the system’s response is fast enough to keep DC bus voltageat a set point. The voltage oscillation during the transient compared to simulation results is higherbecause the simulation is considered an ideal system where there is no acquisition problem and noises.However, in practice, as it can be noticed from Figure 10b, the DC bus has much more noise andoscillation with respect to the simulation one. It is possible to notice the difference between simulationand experimental results in Figure 10b, because, in this case, the voltage probe has higher 30–40 MHzbandwidth. The results are meant to show steady state performance, and the transient behavior wasnot focused on in this paper.

5. Conclusions

Simple MPCC, utilizing advantages of the conventional PI controller, has been designed for asingle phase bidirectional converter. With the proposed PI-MPCC controller variable, the switchingfunction of the original MPC is improved. Moreover, the optimization cost function can be easilyadopted to the controller based on the application. The converter is the basic component of mostrenewable energy types and a major part of the micro grid and smart grid concept. The control methodhas been tested both in MATLAB simulation environment and with a 3 kW experimental prototype.Results show that the inverter has very good transient and steady state performance. The system hasbeen tested both in grid connected and island operation mode. The inverter simulation model matchesvery well to the prototype performance.

Acknowledgments: The research leading to these results has received funding from the Italian Ministry ofEconomic Development (project Smart Domo Grid).

Author Contributions: In this research activity, all the authors were equally involved in the system analysis anddesign, simulation, results analysis and discussion, and manuscript preparation. All authors have approved thesubmitted manuscript.

Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations

The following abbreviations are used in this manuscript:

RES Renewable Energy SourcesDR Distributed ResourcesMG Micro GridsDG Distributed GenerationUPS Uninterruptible Power SuppliesPWM Pulse Width ModulationVOC Voltage Oriented ControlDPC Direct Power ControlDSP Digital Signal ProcessorsFPGA Field-Programmable Gate ArrayMPC Model Predictive ControlPI Proportional IntegralMPCC Model Predictive Current ControllerPI-MPCC PI and MPCCKVL Kirchhoff’s Voltage LawKCL Kirchhoff’s Current LawVSI Voltage Source InverterCCCV Constant Current Constant VoltageDCM Discontinuous Current operation ModeTI Texas InstrumentsLTV Linear Time Varying

Energies 2016, 9, 233 13 of 14

References

1. Bose, B.K. Global warming: Energy, environmental pollution, and the impact of power electronics.IEEE Ind. Electron. Mag. 2010, 4, 6–17. [CrossRef]

2. Karimi-Ghartemani, M. Universal integrated synchronization and control for single-phase DC/ACconverters. IEEE Trans. Power Electron. 2015, 30, 1544–1557. [CrossRef]

3. Accetta, G.; Della Giustina, D.; D’Antona, G.; Faranda, R.; Zanini, S. SmartDomoGrid: Reference architectureand use case analyses for a grid-customer interaction. In Proceedings of the 4th IEEE European InnovativeSmart Grid Technologies (ISGT) Conference, Copenhagen, Danmark, 6–9 October 2013; pp. 1–4.

4. Accetta, G.; D’Antona, G.; Della Giustina, D.; Hafezi, H.; Faranda, R. Open UPQC: A possible solution forcustomer power quality improvement. Shunt unit analysis. In Proceedings of the 16th IEEE InternationalConfrence on Harmonics and Quality of Power, ICHQP, Bucharest, Romania, 25–28 May 2014; pp. 596–600.

5. Kawabata, T.; Higashino, S. Parallel operation of voltage source inverters. IEEE Trans. Ind. Appl. 1988, 24,281–287. [CrossRef]

6. Walling, R.A.; Saint, R.; Dugan, R.C.; Burke, J.; Kojovic, L.A. Summary of distributed resources impact onpower delivery systems. IEEE Trans. Power Del. 2008, 23, 1636–1644. [CrossRef]

7. Cortes, P.; Ortiz, G.; Yuz, J.I.; Rodriguez, J.; Vazquez, S.; Franquelo, L.G. Model predictive control of aninverter with output LC filter for UPS applications. IEEE Trans. Ind. Electron. 2009, 56, 1875–1883. [CrossRef]

8. Carrasco, J.M.; Franquelo, L.G.; Bialasiewicz, J.T.; Galvan, E.; PortilloGuisado, R.C.; Prats, M.A.M.; Leon, J.I.;Moreno-Alfonso, N. Power-electronic systems for the grid integration of renewable energy sources: A survey.IEEE Trans. Ind. Electron. 2006, 53, 1002–1016. [CrossRef]

9. Blaabjerg, F.; Teodorescu, R.; Liserre, M.; Timbus, A.V. Overview of control and grid synchronization fordistributed power generation systems. IEEE Trans. Ind. Electron. 2006, 53, 1398–1409. [CrossRef]

10. Parvez Akter, M.D.; Mekhilef, S.; Nadia, M.L.T.; Akagi, H. Model predictive control of bidirectional AC-DCconverter for energy storage system. J. Electr. Eng. Technol. 2015, 10, 165–175. [CrossRef]

11. Kouro, S.; Cortés, P.; Vargas, R.; Ammann, U.; Rodríguez, J. Model predictive control—A simple andpowerful method to control power converters. IEEE Trans. Ind. Appl. 2009, 56, 1826–1838. [CrossRef]

12. Kojima, M.; Hirabayashi, K.; Kawabata, Y.; Ejiogu, E.C.; Kawabata, T. Novel vector control system usingdeadbeat-controlled PWM inverter with output LC filter. IEEE Trans. Ind. Appl. 2004, 40, 162–169. [CrossRef]

13. Mattavelli, P. An improved deadbeat control for UPS using disturbance observers. IEEE Trans. Ind. Electron.2005, 52, 206–212. [CrossRef]

14. Loh, P.C.; Newman, M.J.; Zmood, D.N.; Holmes, D.G. A comparative analysis of multiloop voltage regulationstrategies for single and three-phase UPS systems. IEEE Trans. Power Electron. 2003, 18, 1176–1185.

15. Buso, S.; Fasolo, S.; Mattavelli, P. Uninterruptible power supply multiloop control employing digitalpredictive voltage and current regulators. IEEE Trans. Ind. Appl. 2001, 37, 1846–1854. [CrossRef]

16. Marwali, M.N.; Keyhani, A. Control of distributed generation systems—Part I: Voltages and currents control.IEEE Trans. Power Electron. 2004, 19, 1541–1550. [CrossRef]

17. Escobar, G.; Mattavelli, P.; Stankovic, A.M.; Valdez, A.A.; Leyva-Ramos, J. An adaptive control for UPSto compensate unbalance and harmonic distortion using a combined capacitor/load current sensing.IEEE Trans. Ind. Electron. 2007, 54, 839–847. [CrossRef]

18. Lee, J.H. Model predictive control: Review of the three decades of development. Int. J. Control Autom. Syst.2011, 9, 415–424. [CrossRef]

19. D’Antona, G.; Faranda, R.; Hafezi, H.; Bugliesi, M. Experiment on bidirectional single phase converterapplying simple model predictive control. In Proceedings of the 2015 IEEE 15th International Conference onEnvironment and Electrical Engineering (EEEIC), Rome, Italy, 10–13 June 2015; pp. 1019–1024.

20. Rodriguez, J.; Kazmierkowski, M.; Espinoza, J.; Zanchetta, P.; Abu-Rub, H.; Young, H.; Rojas, C.A. State ofthe art of finite control set model predictive control in power electronics. IEEE Trans. Power Electron. 2013, 9,1003–1016. [CrossRef]

21. Cortes, P.; Wilson, A.; Kouro, S.; Rodriguez, J.; Abu-Rub, H. Model predictive control of multilevel cascadedH-bridge inverters. IEEE Trans. Ind. Electron. 2010, 57, 2691–2699. [CrossRef]

22. Vahedi, H.; Al-Haddad, K.; Labbe, P.-A.; Rahmani, S. Cascaded multilevel inverter with multicarrier PWMtechnique and voltage balancing feature. In Proceedings of the IEEE 23rd International Symposium onIndustrial Electronics (ISIE), Istanbul, Turkey, 1–4 June 2014; pp. 2155–2160.

Energies 2016, 9, 233 14 of 14

23. Rivera, M.; Wilson, A.; Rojas, C.A.; Rodriguez, J.; Espinoza, J.R.; Wheeler, P.W.; Empringham, L.A comparative assessment of model predictive current control and space vector modulation in a directmatrix converter. IEEE Trans. Ind. Electron. 2013, 60, 578–588. [CrossRef]

24. Rmili, L.; Rahmani, S.; Vahedi, H.; Al-Haddad, K. A comprehensive analysis of matrix converters:Bidirectional switch, direct topology, modeling and control. In Proceedings of the IEEE 23rd InternationalSymposium on Industrial Electronics (ISIE), Istanbul, Turkey, 1–4 June 2014; pp. 313–318.

25. Cortes, P.; Rodriguez, J.; Silva, C.; Flores, A. Delay compensation in model predictive current control of athree-phase inverter. IEEE Trans. Ind. Electron. 2012, 59, 1323–1325. [CrossRef]

26. Xia, C.; Wang, M.; Song, Z.; Liu, T. Robust model predictive current control of three-phase voltage sourcePWM rectifier with online disturbance observation. IEEE Trans. Ind. Inf. 2012, 8, 459–471. [CrossRef]

27. Hafezi, H.; Akpinar, E.; Balikci, A. Cascade PI Controller for Single-phase STATCOM. In Proceedings of the16th International Power Electronics and Motion Control Confrence and Exposition, PEMC, Antalya, Turkey,21–24 September 2014; pp. 127–132.

28. Carmeli, M.S.; Castelli Dezza, F.; Faranda, R.; Marchegiani, G.; Mauri, M. Universal digital controller forpower quality and distributed generation systems. In Proceedings of the International Symposium onPower Electronics, Electrical Drives, Automation and Motion, SPEEDAM, Taormina, Italy, 23–26 May 2006;pp. 508–513.

29. Mohan, N.; Undeland, T.M.; Robbins, W.P. Power Electronics: Converters, Applications, and Design, 2nd ed.;Wiley: New York, NY, USA, 1995.

30. Ho, W.-J.; Lio, J.-B.; Feng, W.-S. Economic UPS structure with phase-controlled battery charger andinput-power-factor improvement. IEE Proc. Electr. Power Appl. 1997, 144, 221–226. [CrossRef]

31. Castelli Dezza, F.; Faranda, R.; Mazzucco, I.; Redi, P.; Tironi, E. An interface converter for DG/storage systemable to improve Power Quality of the load. In Proceedings of the IEEE Power Engineering Society GeneralMeeting PES, Montreal, QB, Canada, 18–22 June 2006.

32. D’Antona, G.; Della Giustina, D.; Faranda, R.; Hafezi, H. Open UPQC Power Quality Manager withinDistributed Generation Systems. In Proceedings of the IEEE 10th International Symposium on Diagnosticsfor Electrical Machines, Power Electronics and Drives (SDEMPED), Guarda, Portugal, 1–4 September 2015;pp. 501–507.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open accessarticle distributed under the terms and conditions of the Creative Commons by Attribution(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).